Fusion modeling for predicting the impact of in-vivo liver motion on HIFU therapies

Here we report in-vivo validation of a fusion modeling method which associates real in-vivo motion data acquired from ultrasound speckle tracking imaging and data from a multi-scale modeling tool. This method aims at estimating the impact of in-vivo liver motion on High Intensity Focused Ultrasound (HIFU) treatments by quantifying thermal lesion deformations. A Dice Similarity Coefficient (DSC) was calculated between lesion volumes obtained with and without organ motions. To validate the model predictions, fusion modeling results were compared to in-vivo intraoperative HIFU treatments performed in pig liver during respiration with 2 different ultrasound focusing strategies: (i) a spherical HIFU transducer used to juxtapose 7 × 7 single millimetric cigar-shaped lesions; (ii) a toroidal-shaped HIFU device developed for the treatment of Liver Metastases and used to create one single centimetric lesion. Liver motions were mainly encountered in the cranial-caudal direction with a magnitude ranging 8-12 mm. Fusion modeling indicated that although a millimetric cigar-shaped lesion was stretched drastically and split in the tissues (DSC <; 50%), centimetric ablations could be achieved intraoperatively without motion correction after 49 lesion juxtapositions in 8 minutes (DSC > 75%). Without motion correction, the toroidal-shaped HIFU strategy allowed generating large homogeneous lesions more than 10 times faster (6 cm3/min). Although the lesion main diameter was stretched by <;15%, liver motion magnitudes observed in this study did not prevent from maintaining acceptable treatment accuracy (DSC > 75%). Overall, fusion modeling of in-vivo HIFU treatments during respiration were well fitted to the observations made experimentally. The fusion between numerical and in-vivo data provided realistic results which could be useful for optimizing HIFU focusing strategies.